Convolutional Neural Networks

Project: Write an Algorithm for a Dog Identification App


In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!

Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the Jupyter Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.

The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this Jupyter notebook.


Why We're Here

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Write your Algorithm
  • Step 6: Test Your Algorithm

Step 0: Import Datasets

Make sure that you've downloaded the required human and dog datasets:

  • Download the dog dataset. Unzip the folder and place it in this project's home directory, at the location /dogImages.

  • Download the human dataset. Unzip the folder and place it in the home directory, at location /lfw.

Note: If you are using a Windows machine, you are encouraged to use 7zip to extract the folder.

In the code cell below, we save the file paths for both the human (LFW) dataset and dog dataset in the numpy arrays human_files and dog_files.

In [1]:
import numpy as np
from glob import glob

# load filenames for human and dog images
human_files = np.array(glob("lfw/*/*"))
dog_files = np.array(glob("dogImages/*/*/*"))

# print number of images in each dataset
print('There are %d total human images.' % len(human_files))
print('There are %d total dog images.' % len(dog_files))
There are 13233 total human images.
There are 8351 total dog images.

Step 1: Detect Humans

In this section, we use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images.

OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory. In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [2]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')

# load color (BGR) image
img = cv2.imread(human_files[0])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Number of faces detected: 1

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [3]:
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer: (You can print out your results and/or write your percentages in this cell)

In [4]:
from tqdm import tqdm

human_files_short = human_files[:100]
dog_files_short = dog_files[:100]

#-#-# Do NOT modify the code above this line. #-#-#

## TODO: Test the performance of the face_detector algorithm 
## on the images in human_files_short and dog_files_short.

face_human = 0
for human_img in tqdm(human_files_short):
    if face_detector(human_img):
        face_human += 1
    else:
        img = cv2.imread(human_img)
        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        print ('Not recognized as a human:')
        plt.imshow(img_rgb)
        plt.show()
        
ratio_human = face_human/len(human_files_short)*100
print ('{}% of the first 100 images in human_files_short have a detected human face'.format(ratio_human))    


face_dog = 0
for dog_img in tqdm(dog_files_short):
    if face_detector(dog_img):
        img = cv2.imread(dog_img)
        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        faces = face_cascade.detectMultiScale(gray)
        for (x,y,w,h) in faces:
            cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        print('\n Human detected in the pictures:')
        plt.imshow(img_rgb)
        plt.show()
        face_dog += 1 
        
ratio_dog = face_dog/len(dog_files_short)*100
print ('{}%  of the first 100 images in dog_files_short have a detected human face'.format(ratio_dog))
  0%|                                                                                          | 0/100 [00:00<?, ?it/s]
Not recognized as a human:
 25%|████████████████████▎                                                            | 25/100 [00:00<00:02, 34.32it/s]
Not recognized as a human:
 37%|█████████████████████████████▉                                                   | 37/100 [00:00<00:01, 37.99it/s]
Not recognized as a human:
 46%|█████████████████████████████████████▎                                           | 46/100 [00:01<00:01, 37.95it/s]
Not recognized as a human:
100%|████████████████████████████████████████████████████████████████████████████████| 100/100 [00:01<00:00, 51.06it/s]
96.0% of the first 100 images in human_files_short have a detected human face
 14%|███████████▎                                                                     | 14/100 [00:00<00:06, 12.58it/s]
 Human detected in the pictures:
 23%|██████████████████▋                                                              | 23/100 [00:01<00:05, 13.74it/s]
 Human detected in the pictures:
 27%|█████████████████████▊                                                           | 27/100 [00:02<00:05, 12.27it/s]
 Human detected in the pictures:
 31%|█████████████████████████                                                        | 31/100 [00:02<00:07,  9.21it/s]
 Human detected in the pictures:
 37%|█████████████████████████████▉                                                   | 37/100 [00:03<00:07,  8.54it/s]
 Human detected in the pictures:
 Human detected in the pictures:
 39%|███████████████████████████████▌                                                 | 39/100 [00:04<00:10,  6.00it/s]
 Human detected in the pictures:
 43%|██████████████████████████████████▊                                              | 43/100 [00:06<00:16,  3.48it/s]
 Human detected in the pictures:
 57%|██████████████████████████████████████████████▏                                  | 57/100 [00:07<00:05,  8.46it/s]
 Human detected in the pictures:
 62%|██████████████████████████████████████████████████▏                              | 62/100 [00:08<00:04,  8.06it/s]
 Human detected in the pictures:
 67%|██████████████████████████████████████████████████████▎                          | 67/100 [00:08<00:03, 10.50it/s]
 Human detected in the pictures:
 72%|██████████████████████████████████████████████████████████▎                      | 72/100 [00:09<00:02, 13.24it/s]
 Human detected in the pictures:
 76%|█████████████████████████████████████████████████████████████▌                   | 76/100 [00:09<00:02,  8.20it/s]
 Human detected in the pictures:
 Human detected in the pictures:
 80%|████████████████████████████████████████████████████████████████▊                | 80/100 [00:10<00:02,  6.83it/s]
 Human detected in the pictures:
 Human detected in the pictures:
 86%|█████████████████████████████████████████████████████████████████████▋           | 86/100 [00:12<00:02,  6.27it/s]
 Human detected in the pictures:
 96%|█████████████████████████████████████████████████████████████████████████████▊   | 96/100 [00:13<00:00,  6.36it/s]
 Human detected in the pictures:
100%|████████████████████████████████████████████████████████████████████████████████| 100/100 [00:13<00:00,  7.15it/s]
18.0%  of the first 100 images in dog_files_short have a detected human face

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.

In [ ]:
### (Optional) 
### TODO: Test performance of another face detection algorithm.
### Feel free to use as many code cells as needed.

Step 2: Detect Dogs

In this section, we use a pre-trained model to detect dogs in images.

Obtain Pre-trained VGG-16 Model

The code cell below downloads the VGG-16 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories.

In [4]:
import torch
import torchvision.models as models

# define VGG16 model
VGG16 = models.vgg16(pretrained=True)

# check if CUDA is available
use_cuda = torch.cuda.is_available()

print(use_cuda)

# move model to GPU if CUDA is available
if use_cuda:
    VGG16 = VGG16.cuda()
True

Given an image, this pre-trained VGG-16 model returns a prediction (derived from the 1000 possible categories in ImageNet) for the object that is contained in the image.

(IMPLEMENTATION) Making Predictions with a Pre-trained Model

In the next code cell, you will write a function that accepts a path to an image (such as 'dogImages/train/001.Affenpinscher/Affenpinscher_00001.jpg') as input and returns the index corresponding to the ImageNet class that is predicted by the pre-trained VGG-16 model. The output should always be an integer between 0 and 999, inclusive.

Before writing the function, make sure that you take the time to learn how to appropriately pre-process tensors for pre-trained models in the PyTorch documentation.

In [5]:
from PIL import Image
import torchvision.transforms as transforms

import torch.nn.functional as F
from collections import OrderedDict
import numpy as np
import torch
from torch import nn
from torch import optim
import json
from PIL import Image
import matplotlib.pyplot as plt
import seaborn as sb
import pandas as pd

# Set PIL to be tolerant of image files that are truncated.
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True


def process_image(image):
    ''' Scales, crops, and normalizes a PIL image for a PyTorch model,
        returns an Numpy array
    '''
    
    # TODO: Process a PIL image for use in a PyTorch model
    size = 256, 256
    new_size = 224
    image = Image.open(image)
    image.thumbnail(size)
    image = image.crop((0, 0, new_size, new_size))
    np_image = np.array(image)
    np_image = np_image / 255
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])
    np_image = (np_image - mean) / std
    np_image = np_image.transpose((2, 0, 1))

    return np_image



def VGG16_predict(img_path, model, topk):
    '''
    Use pre-trained VGG-16 model to obtain index corresponding to 
    predicted ImageNet class for image at specified path
    
    Args:
        img_path: path to an image
        
    Returns:
        Index corresponding to VGG-16 model's prediction
    '''
    
    ## TODO: Complete the function.
    ## Load and pre-process an image from the given img_path
    ## Return the *index* of the predicted class for that image
    
    im = process_image(img_path)
    
    if torch.cuda.is_available():
        input = torch.FloatTensor(im).cuda()
    else:
        input = torch.FloatTensor(im)
    
    input.unsqueeze_(0)
    output = model.forward(input)
    result = F.softmax(output.data, dim=1)  # Alternative method: result = torch.exp(output)
    probs, classes = torch.topk(result, topk)
    probs = probs.data.cpu().numpy()[0]
    classes = classes.data.cpu().numpy()[0]

    return probs, classes   # predicted class index
In [7]:
probs, classes = VGG16_predict("American_eskimo_dog_00415.jpg", VGG16, 5)
print(probs, classes)
[0.7117104  0.16085659 0.11470189 0.00653902 0.00146116] [258 270 279 259 248]

First element in classes (258 = Samoyed) has the highest probability. Samoyed and American Eskimo are very similar dogs. VGG16 seems to be working quite well with dog detection

(IMPLEMENTATION) Write a Dog Detector

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained VGG-16 model, we need only check if the pre-trained model predicts an index between 151 and 268 (inclusive).

Use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [6]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    ## TODO: Complete the function.
    j = 0
    probs, classes = VGG16_predict(img_path, VGG16, 5)
    dog = 0
    
    for i in classes:
        if 151 <= i <= 268 and probs[j] > 0.51:
            dog = 1
            # print ("There is a dog in the picture. Number {} in imagenet 1000 labels dictionary".format(i))
            j += 1
            
    return dog # true/false (1/0)

(IMPLEMENTATION) Assess the Dog Detector

Question 2: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer:

In [9]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.

face_dog = 0
for dog_img in dog_files_short:
    if dog_detector(dog_img) == 1:
        face_dog += 1
    else:
        img = cv2.imread(dog_img)
        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        print ('Not recognized as a dog:')
        plt.imshow(img_rgb)
        plt.show()
        
ratio_dog = face_dog/len(dog_files_short)*100
print ('{}% of the first 100 images in dog_files_short have a detected dog'.format(ratio_dog))    


face_human = 0
for human_img in human_files_short:
    if dog_detector(human_img) == 1:
        img = cv2.imread(human_img)
        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        faces = face_cascade.detectMultiScale(gray)
        for (x,y,w,h) in faces:
            cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
        img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        print('\n Dog detected in the pictures:')
        plt.imshow(img_rgb)
        plt.show()
        face_human += 1 
        
ratio_human = face_human/len(human_files_short)*100
print ('{}%  of the first 100 images in human_files_short have a detected dog'.format(ratio_human))
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
Not recognized as a dog:
79.0% of the first 100 images in dog_files_short have a detected dog

 Dog detected in the pictures:
1.0%  of the first 100 images in human_files_short have a detected dog

We suggest VGG-16 as a potential network to detect dog images in your algorithm, but you are free to explore other pre-trained networks (such as Inception-v3, ResNet-50, etc). Please use the code cell below to test other pre-trained PyTorch models. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.

In [ ]:
### (Optional) 
### TODO: Report the performance of another pre-trained network.
### Feel free to use as many code cells as needed.

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 10%. In Step 4 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have trouble distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador Black Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset

Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dogImages/train, dogImages/valid, and dogImages/test, respectively). You may find this documentation on custom datasets to be a useful resource. If you are interested in augmenting your training and/or validation data, check out the wide variety of transforms!

In [7]:
import os
from torchvision import datasets
import torchvision.transforms as transforms

### TODO: Write data loaders for training, validation, and test sets
## Specify appropriate transforms, and batch_sizes

data_dir = 'dogImages'
train_dir = data_dir + '/train'
valid_dir = data_dir + '/valid'
test_dir = data_dir + '/test'

# Transforms for the training, validation, and testing sets
train_transforms = transforms.Compose([transforms.Resize(224), transforms.CenterCrop(224), transforms.ToTensor(),
                                      transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])

valid_transforms = transforms.Compose([transforms.Resize(224), transforms.CenterCrop(224), transforms.ToTensor(),
                                      transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])

test_transforms = transforms.Compose([transforms.Resize(224), transforms.CenterCrop(224), transforms.ToTensor(),
                                      transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])]) 

# Load the datasets with ImageFolder
train_data = datasets.ImageFolder(train_dir, transform=train_transforms)
valid_data = datasets.ImageFolder(valid_dir, transform=valid_transforms)
test_data = datasets.ImageFolder(test_dir, transform=test_transforms)

# Using the image datasets and the trainforms, define the dataloaders
trainloader = torch.utils.data.DataLoader(train_data, batch_size=10, shuffle=True)
validloader = torch.utils.data.DataLoader(valid_data, batch_size=10, shuffle=True)
testloader = torch.utils.data.DataLoader(test_data, batch_size=10)

loaders_scratch = {
    'train': trainloader,
    'valid': validloader,
    'test': testloader
}

Question 3: Describe your chosen procedure for preprocessing the data.

  • How does your code resize the images (by cropping, stretching, etc)? What size did you pick for the input tensor, and why?
  • Did you decide to augment the dataset? If so, how (through translations, flips, rotations, etc)? If not, why not?

Answer:

All Pytorch pre-trained models expect input images normalized in the same way, i.e. mini-batches of 3-channel RGB images of shape (3 x H x W), where H and W are expected to be at least 224. The images have to be loaded into a range of [0, 1] and then normalized using mean = [0.485, 0.456, 0.406] and std = [0.229, 0.224, 0.225].

I´ve thus applied all this transformations in order to generate a dataloader suitable for both pre-trained models (when using transfer learning) and new models.

Dataset augmentation transformations such as random scaling, cropping, and flipping help the network generalize leading to better performance. The validation and testing sets are anyway used to measure the model's performance on data it hasn't seen yet. For this reason no scaling or rotation transformations should be applied, but it´s anyway needed to resize and then crop the images to the appropriate size. I´ve done just that on the full dataset.

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. Use the template in the code cell below.

In [11]:
import torch.nn as nn
import torch.nn.functional as F

# PyTorch libraries and modules
from torch.autograd import Variable
from torch.nn import Linear, ReLU, CrossEntropyLoss, Sequential, Conv2d, MaxPool2d, Module, Softmax, BatchNorm2d, Dropout

# define the CNN architecture
class Net(nn.Module):
    ### TODO: choose an architecture, and complete the class
    def __init__(self):
        super(Net, self).__init__()
        
        ## Define the 2D convolutional layers
        self.conv1 = nn.Conv2d(3, 96, kernel_size=(3, 3), stride=(1, 1), padding=1)
        self.norm2d1 = nn.BatchNorm2d(96)
        self.conv2 = nn.Conv2d(96, 128, kernel_size=(3, 3), stride=(1, 1), padding=1)
        self.conv3 = nn.Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=1)

        # max pooling layer
        self.pool = nn.MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0)
        
        # dropout layer
        self.dropout3 = nn.Dropout(0.5)
        self.dropout4 = nn.Dropout(0.5)
        
        # Hyperparameters for linear layer
        input_size = 256*28*28
        hidden_size = 500
        output_size = 133
        
        self.fc1 = nn.Linear(input_size, hidden_size)
        self.fc2 = nn.Linear(hidden_size, output_size)
        
        
    def forward(self, x):
        ## Define forward behavior
        x = self.pool(F.relu(self.norm2d1(self.conv1(x))))
        x = self.pool(F.relu(self.conv2(x)))
        x = self.pool(F.relu(self.conv3(x)))
        x = self.dropout3(x) 
        
        # flatten image input
        x = x.view(-1, 256*28*28)
        x = F.relu(self.fc1(x))
        x = self.dropout4(x) 
        x = self.fc2(x)
        return x

#-#-# You do NOT have to modify the code below this line. #-#-#

# instantiate the CNN
model_scratch = Net()
print(model_scratch)

# move tensors to GPU if CUDA is available
if use_cuda:
    model_scratch.cuda()
Net(
  (conv1): Conv2d(3, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (norm2d1): BatchNorm2d(96, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv2): Conv2d(96, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (conv3): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
  (pool): MaxPool2d(kernel_size=(2, 2), stride=(2, 2), padding=0, dilation=1, ceil_mode=False)
  (dropout3): Dropout(p=0.5, inplace=False)
  (dropout4): Dropout(p=0.5, inplace=False)
  (fc1): Linear(in_features=200704, out_features=500, bias=True)
  (fc2): Linear(in_features=500, out_features=133, bias=True)
)

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step.

Answer:

I found a 2014 study from the University of Toroto (by Turing Award winners Geoffrey Hinton and Yoshua Bengio): "Dropout: A Simple Way to Prevent Neural Networks from Overtting" In this study they describe the following CNN architecture tested with good results on The Street View House Numbers (SVHN) DataSet (color images of house numbers collected by Google Street View):

"...The best architecture that we found uses three convolutional layers each followed by a max-pooling layer. The convolutional layers have 96, 128 and 256 filters respectively. Each convolutional layer has a 5 × 5 receptive field applied with a stride of 1 pixel. Each max pooling layer pools 3 × 3 regions at strides of 2 pixels. The convolutional layers are followed by two fully connected hidden layers having 2048 units each. All units use the rectified linear activation function. Dropout was applied to all the layers of the network with the probability of retaining the unit being p = (0.9, 0.75, 0.75, 0.5, 0.5, 0.5) for the different layers of the network (going from input to convolutional layers to fully connected layers). In addition, the max-norm constraint with c = 4 was used for all the weights. A momentum of 0.95 was used in all the layers. These hyperparameters were tuned using a validation set. Since the training set was quite large, we did not combine the validation set with the training set for final training. We reported test error of the model that had smallest validation error..."

I thus started developing a similar architecture but I could not reach the expected results. By trial and error I simplified the model to this final one implemented here above. It has the same number of convolutional and max pooling layers, same number of filters (with different dimensions but same strides) and I have applied dropout only on the fully connected layers (still with probability = 0.5).

In order to avoid to run out of CUDA memory I also had to change the dataloaders batch_size parameter to 10 (I started with 64).

This is how I calculated the input size for the fully connected layers:

dimension = (Image_dimension - kernel_dimension + 2 x padding) / stride + 1

First convolutional layer: dimension_1 = (224-3+2*1)/1+1 = 224

First max pooling layer: dimension_2 = (dimension_1-2+2*0)/2+1 = 112

Second convolutional layer dimension_3 = (dimension_2-3+2*1)/1+1 = 112

Second max pooling layer: dimension_4 = (dimension_3-2+2*0)/2+1 = 54

Third convolutional layer: dimension_5 = (dimension_4-3+2*1)/1+1 = 54

Third max pooling layer: dimension_6 = (dimension_5-2+2*0)/2+1 = 28

(IMPLEMENTATION) Specify Loss Function and Optimizer

Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_scratch, and the optimizer as optimizer_scratch below.

In [12]:
import torch.optim as optim
from torch.optim import Adam, SGD

### TODO: select loss function
criterion_scratch = CrossEntropyLoss()

### TODO: select optimizer
optimizer_scratch = SGD(model_scratch.parameters(), lr=0.001, momentum=0.9)
# optimizer_scratch = SGD(model_scratch.parameters(), lr=0.01)

(IMPLEMENTATION) Train and Validate the Model

Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_scratch.pt'.

In [13]:
# the following import is required for training to be robust to truncated images
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True

def train(n_epochs, loaders, model, optimizer, criterion, use_cuda, save_path):
    """returns trained model"""
    # initialize tracker for minimum validation loss
    valid_loss_min = np.Inf 
    
    for epoch in range(1, n_epochs+1):
        # initialize variables to monitor training and validation loss
        train_loss = 0.0
        valid_loss = 0.0
        
        ###################
        # train the model #
        ###################
        model.train()
        for batch_idx, (data, target) in enumerate(loaders['train']):
            # move to GPU
            if use_cuda:
                data, target = data.cuda(), target.cuda()
            ## find the loss and update the model parameters accordingly
            ## record the average training loss, using something like
            ## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
            
            optimizer.zero_grad()
            
            # Forward and backward passes
            output = model(data)
            loss = criterion(output, target)
            loss.backward()
            optimizer.step()
            
            train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
            
        ######################    
        # validate the model #
        ######################
        model.eval()
        for batch_idx, (data, target) in enumerate(loaders['valid']):
            # move to GPU
            if use_cuda:
                data, target = data.cuda(), target.cuda()
            ## update the average validation loss
            output = model(data)
            loss = criterion(output, target)
            valid_loss = valid_loss + ((1 / (batch_idx + 1)) * (loss.data - valid_loss))

            
        # print training/validation statistics 
        print('Epoch: {} \tTraining Loss: {:.6f} \tValidation Loss: {:.6f}'.format(
            epoch, 
            train_loss,
            valid_loss
            ))
        
        ## TODO: save the model if validation loss has decreased
        if valid_loss < valid_loss_min:
            torch.save(model.state_dict(), save_path)
            print('Validation loss decreased ({:.6f} --> {:.6f}).  Saving model ...'
                  .format(valid_loss_min, valid_loss))
            valid_loss_min = valid_loss
            
    # return trained model
    return model
In [14]:
# train the model
model_scratch = train(20, loaders_scratch, model_scratch, optimizer_scratch, criterion_scratch, use_cuda, 'model_scratch.pt')

# load the model that got the best validation accuracy
model_scratch.load_state_dict(torch.load('model_scratch.pt'))
Epoch: 1 	Training Loss: 4.832814 	Validation Loss: 4.596403
Validation loss decreased (inf --> 4.596403).  Saving model ...
Epoch: 2 	Training Loss: 4.516395 	Validation Loss: 4.345673
Validation loss decreased (4.596403 --> 4.345673).  Saving model ...
Epoch: 3 	Training Loss: 4.280096 	Validation Loss: 4.196826
Validation loss decreased (4.345673 --> 4.196826).  Saving model ...
Epoch: 4 	Training Loss: 4.055470 	Validation Loss: 4.102049
Validation loss decreased (4.196826 --> 4.102049).  Saving model ...
Epoch: 5 	Training Loss: 3.704914 	Validation Loss: 4.027818
Validation loss decreased (4.102049 --> 4.027818).  Saving model ...
Epoch: 6 	Training Loss: 3.195608 	Validation Loss: 4.006811
Validation loss decreased (4.027818 --> 4.006811).  Saving model ...
Epoch: 7 	Training Loss: 2.410840 	Validation Loss: 4.160302
Epoch: 8 	Training Loss: 1.554343 	Validation Loss: 4.482401
Epoch: 9 	Training Loss: 0.981062 	Validation Loss: 4.775261
Epoch: 10 	Training Loss: 0.659445 	Validation Loss: 5.268742
Epoch: 11 	Training Loss: 0.436368 	Validation Loss: 5.042777
Epoch: 12 	Training Loss: 0.354865 	Validation Loss: 5.188484
Epoch: 13 	Training Loss: 0.276268 	Validation Loss: 5.515896
Epoch: 14 	Training Loss: 0.225975 	Validation Loss: 5.538548
Epoch: 15 	Training Loss: 0.193036 	Validation Loss: 5.324347
Epoch: 16 	Training Loss: 0.181626 	Validation Loss: 5.235021
Epoch: 17 	Training Loss: 0.133341 	Validation Loss: 5.740794
Epoch: 18 	Training Loss: 0.110759 	Validation Loss: 5.938570
Epoch: 19 	Training Loss: 0.140178 	Validation Loss: 5.489654
Epoch: 20 	Training Loss: 0.113122 	Validation Loss: 5.762940
Out[14]:
<All keys matched successfully>

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 10%.

In [15]:
def test(loaders, model, criterion, use_cuda):

    # monitor test loss and accuracy
    test_loss = 0.
    correct = 0.
    total = 0.

    model.eval()
    for batch_idx, (data, target) in enumerate(loaders['test']):
        # move to GPU
        if use_cuda:
            data, target = data.cuda(), target.cuda()
        # forward pass: compute predicted outputs by passing inputs to the model
        output = model(data)
        # calculate the loss
        loss = criterion(output, target)
        # update average test loss 
        test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))
        # convert output probabilities to predicted class
        pred = output.data.max(1, keepdim=True)[1]
        # compare predictions to true label
        correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())
        total += data.size(0)
            
    print('Test Loss: {:.6f}\n'.format(test_loss))

    print('\nTest Accuracy: %2d%% (%2d/%2d)' % (
        100. * correct / total, correct, total))

# call test function    
test(loaders_scratch, model_scratch, criterion_scratch, use_cuda)
Test Loss: 3.989833


Test Accuracy: 11% (92/836)

Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset

Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dogImages/train, dogImages/valid, and dogImages/test, respectively).

If you like, you are welcome to use the same data loaders from the previous step, when you created a CNN from scratch.

In [8]:
## TODO: Specify data loaders
loaders_transfer = loaders_scratch.copy()

(IMPLEMENTATION) Model Architecture

Use transfer learning to create a CNN to classify dog breed. Use the code cell below, and save your initialized model as the variable model_transfer.

In [9]:
import torchvision.models as models
import torch.nn as nn

## TODO: Specify model architecture 
model_transfer = models.resnet50(pretrained=True)

for param in model_transfer.parameters():
    param.requires_grad = False

model_transfer.fc = nn.Linear(2048, 133, bias=True)

fc_parameters = model_transfer.fc.parameters()

for param in fc_parameters:
    param.requires_grad = True

model_transfer.class_to_idx = train_data.class_to_idx

print(model_transfer)


if use_cuda:
    model_transfer = model_transfer.cuda()
ResNet(
  (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
  (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (relu): ReLU(inplace=True)
  (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
  (layer1): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer2): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): Bottleneck(
      (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer3): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (3): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (4): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (5): Bottleneck(
      (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (layer4): Sequential(
    (0): Bottleneck(
      (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
      (downsample): Sequential(
        (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      )
    )
    (1): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
      (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
      (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)
      (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
      (relu): ReLU(inplace=True)
    )
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
  (fc): Linear(in_features=2048, out_features=133, bias=True)
)

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer:

Main steps:

  • Load the pre-trained Resnet50 network (already trained on a large dataset of images)

  • Define a new, untrained feed-forward network adding new feedforward classifier to the existing convolutional layers.

  • The parameters of the feedforward classifier are appropriately trained, while the parameters of the feature network (Resnet50) are left static.

This is a standard practice and Resnet50 is widely used in transfer learning projects.

(IMPLEMENTATION) Specify Loss Function and Optimizer

Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_transfer, and the optimizer as optimizer_transfer below.

In [15]:
import torch.optim as optim
from torch.optim import Adam, SGD

criterion_transfer = nn.CrossEntropyLoss()
optimizer_transfer = optim.SGD(model_transfer.fc.parameters(), lr=0.001)

(IMPLEMENTATION) Train and Validate the Model

Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_transfer.pt'.

In [10]:
n_epochs = 20

# train the model
model_transfer = train(n_epochs, loaders_transfer, model_transfer, optimizer_transfer, criterion_transfer, use_cuda, 'model_transfer.pt')

# load the model that got the best validation accuracy (uncomment the line below)
model_transfer.load_state_dict(torch.load('model_transfer.pt'))
Epoch: 1 	Training Loss: 4.701534 	Validation Loss: 4.378246
Validation loss decreased (inf --> 4.378246).  Saving model ...
Epoch: 2 	Training Loss: 4.267260 	Validation Loss: 3.873397
Validation loss decreased (4.378246 --> 3.873397).  Saving model ...
Epoch: 3 	Training Loss: 3.874635 	Validation Loss: 3.446469
Validation loss decreased (3.873397 --> 3.446469).  Saving model ...
Epoch: 4 	Training Loss: 3.519276 	Validation Loss: 3.073528
Validation loss decreased (3.446469 --> 3.073528).  Saving model ...
Epoch: 5 	Training Loss: 3.205335 	Validation Loss: 2.740608
Validation loss decreased (3.073528 --> 2.740608).  Saving model ...
Epoch: 6 	Training Loss: 2.935496 	Validation Loss: 2.437935
Validation loss decreased (2.740608 --> 2.437935).  Saving model ...
Epoch: 7 	Training Loss: 2.687978 	Validation Loss: 2.183903
Validation loss decreased (2.437935 --> 2.183903).  Saving model ...
Epoch: 8 	Training Loss: 2.484931 	Validation Loss: 2.029093
Validation loss decreased (2.183903 --> 2.029093).  Saving model ...
Epoch: 9 	Training Loss: 2.290971 	Validation Loss: 1.788817
Validation loss decreased (2.029093 --> 1.788817).  Saving model ...
Epoch: 10 	Training Loss: 2.127159 	Validation Loss: 1.650707
Validation loss decreased (1.788817 --> 1.650707).  Saving model ...
Epoch: 11 	Training Loss: 1.996266 	Validation Loss: 1.519300
Validation loss decreased (1.650707 --> 1.519300).  Saving model ...
Epoch: 12 	Training Loss: 1.871621 	Validation Loss: 1.439600
Validation loss decreased (1.519300 --> 1.439600).  Saving model ...
Epoch: 13 	Training Loss: 1.766269 	Validation Loss: 1.353934
Validation loss decreased (1.439600 --> 1.353934).  Saving model ...
Epoch: 14 	Training Loss: 1.672887 	Validation Loss: 1.260041
Validation loss decreased (1.353934 --> 1.260041).  Saving model ...
Epoch: 15 	Training Loss: 1.582914 	Validation Loss: 1.188604
Validation loss decreased (1.260041 --> 1.188604).  Saving model ...
Epoch: 16 	Training Loss: 1.522261 	Validation Loss: 1.101940
Validation loss decreased (1.188604 --> 1.101940).  Saving model ...
Epoch: 17 	Training Loss: 1.447569 	Validation Loss: 1.094382
Validation loss decreased (1.101940 --> 1.094382).  Saving model ...
Epoch: 18 	Training Loss: 1.399255 	Validation Loss: 1.027737
Validation loss decreased (1.094382 --> 1.027737).  Saving model ...
Epoch: 19 	Training Loss: 1.334607 	Validation Loss: 1.011156
Validation loss decreased (1.027737 --> 1.011156).  Saving model ...
Epoch: 20 	Training Loss: 1.281479 	Validation Loss: 0.960508
Validation loss decreased (1.011156 --> 0.960508).  Saving model ...
Out[10]:
<All keys matched successfully>

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 60%.

In [11]:
def test(loaders, model, criterion, use_cuda):

    # monitor test loss and accuracy
    test_loss = 0.
    correct = 0.
    total = 0.

    model.eval()
    for batch_idx, (data, target) in enumerate(loaders['test']):
        # move to GPU
        if use_cuda:
            data, target = data.cuda(), target.cuda()
        # forward pass: compute predicted outputs by passing inputs to the model
        output = model(data)
        # calculate the loss
        loss = criterion(output, target)
        # update average test loss 
        test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))
        # convert output probabilities to predicted class
        pred = output.data.max(1, keepdim=True)[1]
        # compare predictions to true label
        correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())
        total += data.size(0)
            
    print('Test Loss: {:.6f}\n'.format(test_loss))

    print('\nTest Accuracy: %2d%% (%2d/%2d)' % (
        100. * correct / total, correct, total))


test(loaders_transfer, model_transfer, criterion_transfer, use_cuda)
Test Loss: 0.959283


Test Accuracy: 81% (684/836)

(IMPLEMENTATION) Predict Dog Breed with the Model

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan hound, etc) that is predicted by your model.

In [16]:
# load the model that got the best validation accuracy (uncomment the line below)
model_transfer.load_state_dict(torch.load('model_transfer.pt'))

test(loaders_transfer, model_transfer, criterion_transfer, use_cuda) # Just checking that the model has been loaded correctly
Test Loss: 0.959283


Test Accuracy: 81% (684/836)
In [17]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.

data_transfer = loaders_transfer.copy()

# list of class names by index, i.e. a name can be accessed like class_names[0]
class_names = [item[4:].replace("_", " ") for item in data_transfer['train'].dataset.classes]


def process_image(image2):
        ''' Scales, crops, and normalizes a PIL image for a PyTorch model,
            returns an Numpy array
        '''
        size = 256, 256
        new_size = 224
        image2 = Image.open(image2)
        image2.thumbnail(size)
        image2 = image2.crop((0, 0, new_size, new_size))
        np_image2 = np.array(image2)
        np_image2 = np_image2 / 255
        mean = np.array([0.485, 0.456, 0.406])
        std = np.array([0.229, 0.224, 0.225])
        np_image2 = (np_image2 - mean) / std
        np_image2 = np_image2.transpose((2, 0, 1))

        return np_image2



def predict_breed_transfer(img_path2, model2, topk2):
    # load the image and return the predicted breed
    img2 = Image.open(img_path2)
    plt.imshow(img2)
    plt.show()
    
    im2 = process_image(img_path2)
    
    if torch.cuda.is_available():
        input = torch.FloatTensor(im2).cuda()
    else:
        input = torch.FloatTensor(im2)

    input.unsqueeze_(0)
    output2 = model2.forward(input)
    result2 = F.softmax(output2.data, dim=1)  # Alternative method: result = torch.exp(output)
    probs2, classes2 = torch.topk(result2, topk2)
    probs2 = probs2.data.cpu().numpy()[0]
    classes2 = classes2.data.cpu().numpy()[0]

    # predicted_classes = [classname for classname, val in model.class_to_idx.items() if val in classes]
    
    prob = 0
    k = 0
    breed = 0
    for i in probs2:
        if i >= prob:
            prob = i
            breed = classes2[k]
        k+=1

    return probs2, classes2, breed, class_names[breed]
In [18]:
for file in np.hstack((dog_files[60:65])):
    probs2, classes2, breed, class_names[breed] = predict_breed_transfer(file, model_transfer, 1)
    print("With probability {}% the dog in the picture is: {}".format(probs2*100, class_names[breed]))
With probability [58.652985]% the dog in the picture is: American staffordshire terrier
With probability [62.111004]% the dog in the picture is: American staffordshire terrier
With probability [61.240685]% the dog in the picture is: American staffordshire terrier
With probability [4.707436]% the dog in the picture is: Boykin spaniel
With probability [17.257223]% the dog in the picture is: Irish water spaniel

Step 5: Write your Algorithm

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 4 to predict dog breed.

Some sample output for our algorithm is provided below, but feel free to design your own user experience!

Sample Human Output

(IMPLEMENTATION) Write your Algorithm

In [19]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.

def run_app(img_path, model, topk):
    ## handle cases for a human face, dog, and neither
    
    if face_detector(img_path) > 0:
        probs, classes, breed, class_names[breed] = predict_breed_transfer(img_path, model, topk)
        print("Human found in the picture resembling a dog ({} with probability {}%)".format(class_names[breed], probs2*100))
        
    elif dog_detector(img_path):
        probs, classes, breed, class_names[breed] = predict_breed_transfer(img_path, model, topk)
        print("With probability {}% the dog in the picture is: {}".format(probs2*100, class_names[breed]))
        
    else:
        img = Image.open(img_path)
        plt.imshow(img)
        plt.show()
        print('Error: neither human nor dog detection')
In [20]:
## suggested code, below
for file in np.hstack((human_files[:5], dog_files[:5])):
    run_app(file, model_transfer, 1)
Human found in the picture resembling a dog (French bulldog with probability [17.257223]%)
Human found in the picture resembling a dog (American water spaniel with probability [17.257223]%)
Error: neither human nor dog detection
Human found in the picture resembling a dog (Australian shepherd with probability [17.257223]%)
Human found in the picture resembling a dog (Dachshund with probability [17.257223]%)
With probability [17.257223]% the dog in the picture is: Affenpinscher
With probability [17.257223]% the dog in the picture is: Affenpinscher
Error: neither human nor dog detection
With probability [17.257223]% the dog in the picture is: Affenpinscher
With probability [17.257223]% the dog in the picture is: Affenpinscher

Step 6: Test Your Algorithm

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

Answer: (Three possible points for improvement)

The algorithm works well and creates funny suggestions for the dog´s breeds that are mostly resamling the human pictures. There are still anyway some imprecisions (as expected) when classifing dog´s pictures.

Three possible points for improvement could be:

  • expanding the training and validation datasets with more pictures and re-train the network. The higher the quantity of data the better the outcome of the training.

  • including within the dataset also "difficult" pictures where the dog is not in a natural position or where other disturbance factors may lead to a wrong calssification. Edge cases are of cource the most difficult to be identified.

  • training another feedforward classifier (using a different pre-trained model and transfer learning) and then run the two networks in parallel comparing their outcomes prior issuing a final classification result.

In [21]:
## TODO: Execute your algorithm from Step 6 on
## at least 6 images on your computer.
## Feel free to use as many code cells as needed.

import os
from PIL import Image

for img_file in os.listdir('./HUMANS'):
    img_path = os.path.join('./HUMANS', img_file)
    run_app(img_path, model_transfer, 1)
    
for img_file in os.listdir('./DOGS'):
    img_path = os.path.join('./DOGS', img_file)
    run_app(img_path, model_transfer, 1)
Human found in the picture resembling a dog (Neapolitan mastiff with probability [17.257223]%)
Human found in the picture resembling a dog (English toy spaniel with probability [17.257223]%)
Human found in the picture resembling a dog (English toy spaniel with probability [17.257223]%)
Error: neither human nor dog detection
With probability [17.257223]% the dog in the picture is: French bulldog
Error: neither human nor dog detection
In [ ]: